the dynamic nature of apomixis in the angiosperms

14 Pages·2007·0.24 MB·English

Int.J.PlantSci.169(1):169–182.2008.
(cid:2)2008byTheUniversityofChicago.Allrightsreserved.
1058-5893/2008/16901-0014$15.00 DOI:10.1086/523369
THE DYNAMIC NATURE OF APOMIXIS IN THE ANGIOSPERMS
Jeannette Whitton,1,* ChristopherJ. Sears,* EricJ. Baack,*and SarahP.Ottoy
*DepartmentofBotany,UniversityofBritishColumbia,Vancouver,BritishColumbiaV6T1Z4,Canada;and
yDepartmentofZoology,UniversityofBritishColumbia,Vancouver,BritishColumbiaV6T1Z4,Canada
Apomixis,theasexualproductionofseed,isatraitestimatedtooccurinfewerthan1%offloweringplant
species,withanunevendistributionamonglineages.Inthepastdecade,targetedresearcheffortshaveaimedat
clarifyingthegeneticbasisofapomixis,withthegoalofengineeringorbreedingapomicticcrops.Recentwork
suggestsasimplegeneticbasisforapomixis,butitalsoindicatesthatnaturalpopulationsofapomictsaremuch
morecomplexthanisoftenassumed.Forexample,innature,nearlyallapomictsthatgothroughamegaga-
metophytestage(gametophyticapomicts)arepolyploid,whiletheirsexualrelativesaretypicallydiploid.Al-
thoughpopulationshavebeencharacterizedasobligatelysexualorapomictic,itisincreasinglyclearthatmany
plantpopulationsexhibitsomevariationinreproductivemode.Manyapomictsretainresidualsexualfunction
aspollendonorsandthushavethepotentialtospreadapomixisviamalegametes,therebyincreasingthege-
neticdiversityobservedwithinapomicticpopulations.Here,wesummarizeourcurrentunderstandingofthege-
neticbasisandtransmissionofapomixis.Weuseinsightsfrompreviouscasestudiesandmodelsforthespread
ofasexuality to explore thepotential for establishmentandspreadof apomixis innature.
Keywords:apomixis, asexual reproduction,cost of sex,Crepis,polyploidy.
Introduction apomictic complexes, notably in Taraxacum (van der Hulst
et al. 2000; Verduijn et al. 2004; Meirmans et al. 2006), Hi-
Sexualreproductionisanearlyuniversalcharacteristicofan- eracium (Gadella 1987; Bicknell et al. 2003; Houliston and
giosperms. Despite the overwhelming importance and broad Chapman2004),andCrataegus(MuniyamaandPhipps1979a,
occurrenceofsexuality,mostplantspecies(otherthanannuals) 1979b,1984a,1984b;DickinsonandPhipps1986;Talentand
arecapableofsomeformofasexualpropagation,thoughin Dickinson2005,2007).Here,wefocusonthegenetic,popu-
themajorityofcases,thisremainsanaugmentationtosexual lationgenetic,andecologicalfactorsthataffectthedynamics
reproduction rather than the dominant reproductive mode ofestablishmentandspreadofasexuallineages.Weareespe-
(Richards2003;Silvertown2008).Inanumberofgroups,how- cially interested in the contributions of occasional sexual re-
ever,asexualityhasbecomepredominantthroughevolutionary productiontotheestablishmentandspreadofasexualityand
transitionsfromsexuality.Whereassomeplantshaveevolved specifically in therole thatpollen playsin thiscontext. May-
asexuality throughthedegenerationof sexualstructures infa- nardSmith(1978,p.66)notedthepotentialimportanceofpol-
vor of vegetative means of propagation (Eckert 2002; Silver- len in the spread of apomixis, but this area remained largely
town2008),themorecommonroutetoasexualityisthrough unexploredbeforeworkbyMogie(1992).Sincethistime,sub-
theevolutionofapomixis,theproductionofclonalseedinthe stantialinsightsintothespreadofapomixishavebeengleaned
absenceoffertilization(Richards1986).FirstdescribedbySmith fromtheoreticalandempiricalwork,andwesummarizethese
(1841),apomixisisnowreportedinmorethan300speciesin findingsandidentifyareasforfuturework.
more than 40 angiosperm families (Asker and Jerling 1992).
Thus, evolutionary transitions to asexual reproduction have
occurred repeatedly in flowering plants, with only rare shifts
A Brief Description of Apomictic Phenomena
inthereversedirection(Chapmanetal.2003).
Asexualreproductioninplantshasreceivedsteadyattention
Throughout this review, ‘‘apomixis’’ is used synonymously
fromecologistsandevolutionarybiologistsduringthepastsev-
with‘‘agamospermy’’(Stebbins1950),referringtotheasexual
eraldecades,includingsummariesinclassicworksbyStebbins
productionofseed.Othertermsdescribingapomixisaredefined
(1950),Harper(1977),Grant(1981),andRichards (1986),
in table 1. Apomixis is unknown in gymnosperms, although
among manyothers.Inrecentyears,significantattentionhas
cleavage polyembryony, the cloning or twinning of embryos
been given to determining the genetic basis of apomixis (re-
fromthesamesexualzygote,iswelldocumented(Mogie1992).
viewsinGrimanellietal.2001;Ozias-Akins2006),withan
Thepathwaysbywhichapomicticseedsareproducedaredi-
eye to the development of apomictic crops. Another area of
videdintothreebroadcategories(table1;Stebbins1950):ad-
active research focuses on understanding natural variation in
ventitiousembryony,diplospory,andapospory.Thelattertwo
categories are referred to collectively as gametophytic apo-
1Authorforcorrespondence;e-mail:[email protected].
mixis(Stebbins1950;Grant1981;AskerandJerling1992).In
ManuscriptreceivedMay2007;revisedmanuscriptreceivedAugust2007. eachcase,apomicticembryosarederivedfrommaternalgenetic
169
170 INTERNATIONAL JOURNAL OF PLANTSCIENCES
Table 1
DescriptionofApomicticPhenomenainFloweringPlants
Frequency
Endosperm ofapomictic
Modeofapomixis Originofembryo development Ploidy Pollen Geneticbasis seeds Taxa
Adventitious Somatictissue Requires Usually Necessaryfor Littlestudied; ?? Widelydistributed,
embryony surroundingovule; development diploid sexual onelocus especiallyRutaceae,
sexualovulemust ofsexually reproduction dominant? Celestraceae,and
befertilized fertilizedseed Orchidaceae
Gametophytic
apomixis:
Diplospory Unreduced Usually Rarely Maytransmit Twounlinked High Asteraceae
megaspore autonomous diploid apomixis loci
mothercellgives
risetounreduced
megagametophyte
Apospory Nucellusgivesrise Usually Rarely Necessaryfor Singlelinkage High Poaceaeand
tounreduced pseudogamous diploid endosperm block;1þ Rosaceae
megagametophyte development dominantloci
material,buteachisassociatedwithdifferentprobabilitiesof ized (Bicknell et al. 2003). In the following, we describe the
producingsexualprogeny,differentselectionpressurestomain- major classes of apomicts in more detail and discuss the im-
tainmalefertility,and,consequently,differentexpectedlevels plicationsofeachforpollenmaintenanceandfortheproduc-
ofgeneticdiversitywithinpopulations. tionofmixturesofsexualandasexualseed.
In adventitious embryony, embryos develop from somatic
cells (either from the nucellus or from the integument of the Adventitious Embryony
ovule) rather than from the megagametophyte. In gameto-
phytic apomixis, unreduced megagametophytes are produced Unlike gametophytic apomixis, adventitious embryony has
that subsequently develop into embryos. In this case, the cell the potential to occur in parallel with sexual megagameto-
that gives rise to the megagametophyte can have one of two phyte development (Nygren 1967; Asker and Jerling 1992).
origins.Indiplospory,theunreducedmegagametophyteispro- In most cases that have been examined, the sexual embryo
ducedbythemodificationorcircumventionof meiosisinthe sacappearstodevelopnormally,andpollinationfollowedby
megaspore mothercell—the samecell thatwouldgiverise to double fertilization initiates both sexual embryo and endo-
sexualmegagametophytes.Inapospory,theunreducedmega- sperm development. Once the sexual embryo is initiated, ad-
gametophyte arises through mitotic divisions of a cell of the ditionalsomaticembryosdevelopfromcellsofthenucellusor
nucellus, usually in conjunction with or following degenera- the integuments (Naumova 1993). A large proportion of taxa
tionofthesexualmegagametophyte.Diplosporyandapospory with adventitious embryony can produce multiple embryos per
describe the processes generating the mature megagameto- seed (polyembryony), one of which may be sexual (Richards
phyte; apomixis results when the unfertilized egg undergoes 2003).Insomecases,theseembryoshavebeenshowntocom-
parthenogeneticdevelopmenttoproducetheembryo. petefortheresourcesoftheendosperm(Grant1981;Naumova
Some apomicts require pollen for proper seed maturation 1993),withthepercentageofasexualseedvaryingaccording
(the alternative is called autonomous apomixis; Nygren 1967). to the outcome of the competition. The relative success of
In most of these cases, pollen is necessary for the proper de- sexualversusasexualembryosthereforevariesamongspecies
velopment of the endosperm, with at least one of the pollen and conditions. For example, the percentage of asexual seed
nucleifusingwithatleastoneofthepolarnucleiofthemega- rangesfrom33%inEurekalemonto100%inDancymandarin
gametophyte(Richards1986).Thisphenomenonisknownas amongCitruscultivars(Reutheretal.1968,inGrant1981).
pseudogamy.Adventitiousembryonyisusuallypseudogamous. Numerous embryological observations of at least occa-
Amonggametophyticapomicts,pseudogamyisprevalentamong sional adventitious embryony led some authors to conclude
aposporous apomicts, whereas autonomous endosperm for- that this is the most taxonomically widespread form of apo-
mation is more common with diplospory (Richards 1986). mixis(Naumova1993;Carman1997),despitebeingtheleast
Therequirementforfertilizationoftheendospermselectsfor wellknownintermsofbothitsimportanceinnatureandthe
the maintenance of at least some viable pollen (Noirot et al. genetic basis of apomixis. With the occurrence of sexual and
1997).Apomictswithautonomousendospermformationtend asexualprocessesalongsideeachother,parallelsareoftendrawn
to produce less viable pollen and, in some cases, are male between adventitious embryony and vegetative reproduction
sterile (Meirmans et al. 2006; Thompson and Whitton 2006; because asexuality is often facultative in both cases (Grant
Thompson et al. 2008). In rare cases, apomicts have been 1981;Nogler1984).Itisunclearwhetherthisisafairrepre-
shown to require pollination to stimulate seed development, sentation, however, because so little is known about the fre-
even though neither the embryo nor the endosperm is fertil- quency of asexual seed production in taxa with adventitious
WHITTON ETAL.—APOMIXIS IN ANGIOSPERMS 171
embryony. Furthermore, the phenomena of adventitious apo- phytes (Bicknell et al. 2003). In some cases, however, even if
mixisandvegetativereproductionhaveverydifferentimplica- maturesexualembryosacsareproducedandtheaposporous
tionsfordispersal andfortheproductionofseedscapableof initialsfailtodevelop,sexualembryosmaynotbeviable(Nogler
survivinginhospitableseasons(Richards2003). 1984). Whether this represents anomalies in sexual develop-
Anotherinterestingfeatureofadventitiousapomictsisthat ment that contributed to the selective advantage of apomixis
theyarefrequentlydiploid(orofthesameploidyastheirsex- or reflects a breakdown in sexual reproduction that occurred
ualrelatives),asituationthatisexceedinglyrareamonggame- afterapomixisaroseisunknown.
tophytic apomicts (Asker and Jerling 1992; Koltunow 1993). The distinction between apospory and diplospory is not as
That so little is known about the population biology of ad- sharpasthepreviousdescriptionswouldsuggest.Thisisbest
ventitiousembryony(Naumova1993)limitsourabilitytodraw illustrated in the Rosaceae, where apospory and diplospory
inferencesaboutthisphenomenon. havebeendescribedinthesameindividuals(Muniyammaand
Phipps1984a).Thisintriguingobservationmayarisefromthe
difficultyofdistinguishingsomaticandgenerativecellsinthe
Gametophytic Apomixis
multicellularfemalearchesporium(thepartofthenucellusthat
gives rise to the megaspore mother cell) of many Rosaceae
The two forms of gametophytic apomixis, apospory and
(Asker1980).However,thereareseveralspeciesinwhichboth
diplospory,havebeenstudiedinfargreaterdepththanadven-
diplosporyandaposporyoccur,leadingsomeworkerstospec-
titious embryony and are better understood (Richards 1986;
ulate that these two modes of megagametophyte formation
Mogie 1992). Among the most striking features of gameto-
shareacommongeneticbasis(Mogie1992).
phytic apomixis is its strong correlation with polyploidy: of
examplesinmorethan126genera(Carman1997),lessthana
handfularereportedtoincludediploids(Nogler1984;Mogie
Taxonomic Occurrence of Apomixis
1992; Schranz et al. 2006), and some of these cases are not
supported by recent studies (e.g., Potentilla argentea; Holm
Apomixis occurs sporadically among the ca. 457 angio-
etal.1997).Whereasoddploidylevels(e.g.,triploidy,penta-
sperm families (APG 2003). Carman (1997) lists more than
ploidy) are often reliable predictors of the presence of apo-
330generawithapomixis,withmorethantwo-thirdsofthese
mixis, tetraploidy is the most common ploidy level among
beingtaxawithadventitiousembryony.Thefrequencyofasex-
apomicts(AskerandJerling1992).Explanationsfortheasso-
ualreproductionintaxareportedtoproduceadventitiousem-
ciation between gametophytic apomixis and polyploidy are
bryos is poorly documented; thus, this number provides less
numerousandvaried;theyincludemechanismsthatfocuson
informationabouttheimportanceofasexualityinnaturethan
theecologyofpolyploidsandapomicts(Stebbins1950),thege-
dothenumbersforgametophyticapomicts.Ofca.126genera
netic consequences of polyploidy and apomixis (Lokki 1976;
known to include gametophytic apomicts (Carman 1997),
Manning and Dickson 1986), the genetic basis of apomixis
roughlythree-fourthsofcasesoccurinjustthreefamilies—the
(Mogie 1992), and the shared role of unreduced gamete for-
Rosaceae,Poaceae,andAsteraceae,whichtogethercompriseca.
mation(HarlananddeWet1975).Wereturntothisassociation
15%ofangiospermspecies(Richards1986;AskerandJerling
in‘‘WhatIstheGeneticBasisofApomixis?’’
1992; APG 2003). Estimating the total number of apomictic
speciesisinherentlyproblematicbecauseofthewidedifferences
Diplospory
ofopiniononthetaxonomictreatmentofapomicts(Richards
Diplosporousembryosaretheproductofeitheracomplete 2003),butitisunlikelythatmorethan1%offloweringplant
omission of meiosis oran abnormal meiosis that yields unre- species are substantially apomictic. Predominant mechanisms
duced products, one of which develops mitotically into the ofapomixisalsodifferamongtaxonomicgroups:membersof
unreduced megagametophyte. Because diplosporous megaga- theRosaceaeandPoaceaearemostoftenaposporous,whereas
metophytesarederivedfromthemegasporemothercell, dip- apomictic Asteraceae are commonly diplosporous (table 1).
losporyinterferesdirectlywithsexualreproduction,andthus Reportsofadventitiousembryonyarerareinthesethreefami-
individualsexhibitingdiplosporyaremorelikelytobeobligately lies (Asker and Jerling 1992; Naumova 1993). Adventitious
asexual,assumingthatwhateverisinterferingwithnormalmei- embryonyismostfrequentintheRutaceae,Celastraceae,and
osisis expressed uniformly in all ovules. Diplospory can lead Orchidaceae(Naumova1993).Althoughadventitiousembryony
to genetically variable offspring ifcrossing over or automixis ismostcommonintropicalorsubtropicaltreesandshrubs,ga-
(Thompson and Ritland 2006) occurs before the cessation of metophyticapomixisisdescribedasmostcommonintemper-
meiosis. ate perennial herbs (Asker and Jerling 1992). However, no
surveys of apomixis to date correct for possible correlations
Apospory
betweentaxonomyandgrowthformsorgeographicdistribution.
In aposporous apomicts, the unreduced megagametophyte Regardlessoftheprecisenumbers,thebroadtaxonomicdis-
isderivedfromasomaticcelloftheovule(ratherthanthemega- tribution indicates that apomixis, in all of its various forms,
sporemothercell),whichundergoesmitoticdivisionstoform hasarisenmultipletimes.Italsosuggeststhatthedifferentforms
amegagametophyte.Meiosisinthemegasporemothercellusu- of apomixis either are more likely to arise or are more likely
allyappearsnormal,butproductsmaybegintodegeneratesoon to establish in some groups than in others. What might ac-
aftermeiosisiscompleted(Albertinietal.2001).Aposporous countforsuchpatterns?Differentexplanationshavebeenof-
apomicts have been shown to produce some sexual offspring fered for the distribution of different forms of apomixis. For
through fertilization of reduced and unreduced megagameto- example,ithasrecentlybeenobservedthatpollenlimitationis
172 INTERNATIONAL JOURNAL OF PLANTSCIENCES
moresevereintropicaltaxa,especiallyamongself-incompatible cosegregatedinthemajorityofprogeny,twoindividualsshowed
species and trees (Vamosi et al. 2006), which might selec- signs of apospory without parthenogenesis, supporting the idea
tively favor adventitious embryony as a strategy to augment that at least two loci inherited as a single linkage block con-
reproductive output. However, the taxonomic distribution of trolaposporousapomixis.
apomixismaybedeterminedbyappropriategeneticopportu- Thegeneticcontrolofdiplosporousapomixisiswellstudied
nities rather than being favored by ecological circumstances. in three systems: Erigeron and Taraxacum in the Asteraceae
Members of theRosaceae, Poaceae, andAsteraceaehave been and Tripsacum in the Poaceae (Ozias-Akins 2006). In two of
shown to produce high frequencies of unreduced gametes, es- these cases, control of apomixis involves two unlinked loci
peciallyinhybrids(RamseyandSchemske1998,theirsupple- separately controlling the production of diplosporous mega-
mentary table 1), which might provide repeated opportunities gametophytes and their parthenogenetic development (van
for the evolution of gametophytic apomixis. However, the Dijketal.1999;NoyesandRieseberg2000;Noyesetal.2007).
same is true of other taxa (e.g., Lilliaceae, Convolvulaceae) In Tripsacum, although the two components of apomixis ap-
inwhichapomixisisnotparticularlycommon.Furtherstudy peartoco-occuronasinglelinkagegroup,recombinationbe-
isnecessarytoconfirm—orrefute—thesehypotheses. tweenthemhasbeendetected(Grimanellietal.1988).However,
only in recent work in Erigeron (Noyes et al. 2007) does re-
combination appear tooccur freely, withnearlyequal propor-
What Is the Genetic Basis of Apomixis? tionsofthefourexpectedphenotypiccombinationsfromacross
betweenasexualseedparentandanapomict-sexualhybrid.
Thepromiseandprospectofharnessingapomixisasatool Interesting trends emerge from the genetic data obtained
for use in agriculture and plant breeding have resulted in the thusfar.First,theevidence consistentlysuggeststhatmostof
proliferation of studies aimed at identifying genes that con- the alleles controlling apomixis are dominant. This may not
tributetothecontrolofapomixis,withapomicticgrassesthus be surprising, given the association between apomixis and
far receiving the most attention (Ozias-Akins 2006). Most polyploidy, which, in a tetraploid, would require four copies
genetic studies rely on the production of meiotically reduced ofrecessiveallelestobeexpressed.Mechanistically,dominance
viable pollen in apomicts to study segregation of the trait in suggeststhatapomixisinvolvesanactivetriggerofadventitious
F progeny from crosses with sexually reproducing female embryony,unreducedgameteformation,andparthenogenesis
1
parents.Anumberofexcellentreviewsprovidedetailedsum- ratherthaninactivatingmutationsthatknockoutnormalfunc-
maries of the genetics of apomixis (Grimanelli et al. 2001; tions,whichtendtoberecessive.Thisiscontrarytotheviews
Grossniklausetal.2001;Ozias-Akins2006).Asintheseworks, of Grimanelli et al. (2001), who suggest that apomixis arises
wefocushereonmorerecentstudiesusingmoleculargenetic asaderegulationofsexualfunction.
approaches. A puzzle about the genetics of apomixis raised by Asker
The genetic basis of adventitious embryony has apparently (1980) and others is the requirement for nearly simultaneous
been studied only in Citrus (Garcia et al. 1999). In this case, transitionsintheformationofmegagametophytesandinem-
the3:1ratioofapomictictosexualoffspringsuggeststheac- bryodevelopment.Thechancethattwomutationscausingthese
tionofasinglelocus,withtheapomixisallelebeingdominant. two shifts would occur soon after each other within a small
However, quantitative trait locus (QTL) analysis identifies at nascent apomictic population seems prohibitive; yet, without
least three independently segregating markers that appear to thesimultaneousemergenceofthetwotraits,apomixisseems
be associated with apomixis in this system, with additional unlikely. For example, the development of functional diplo-
lociinvolvedinthecontrolofpolyembryony.TheQTLanalysis sporous megagametophytes without parthenogenesis would
relied on relatively few markers (69) and progeny (50), and lead to embryos with increasing ploidy in each generation,
sotheseconclusionsaresomewhatpreliminary. whereastheconversewouldproducehaploids.Thispuzzleled
Gametophyticapomixisistypicallydissectedintotwotraits, Asker and Jerling (1992) to argue in favor of a single-gene
whichpotentiallyhavedifferentgeneticbases:unreducedmega- mechanism, with one mutation causing both unreduced egg
gametophyteformation (through aposporyordiplospory) and formationandparthenogenesis.Todate,however,severalstud-
parthenogenetic development of the embryo. In the case of ies(citedpreviously)implicatemultipleloci,whereasnostudy
autonomous apomixis, proper endosperm development with- hasyettodemonstratethatonlyasinglegeneisinvolved(al-
outfertilizationrepresentsathirdgeneticchangerequiredfor thoughthismayturnouttobethecaseinsomeoftheexam-
apomixis. Studies of the genetic basis of apospory in seven ples involving a single linkage group). One way around this
species(Poaceae:Pennisetumsquamulatum,Pennisetumciliare, puzzle is suggested by the observation that parthenogenesis
Panicum maximum, Brachiaria sp., Paspalum notatum; Ra- may be triggered by the presence of unreduced megagameto-
nunculaceae: Ranunculus sp.;Asteraceae:Hieraciumsp.) have phytes,atleasttosomeextent.AmongthesegregantsofErigeron
found that aposporous megagametophyte development and examined by Noyes and Rieseberg (2000), offspring inheriting
parthenogenesiscosegregateasasingledominantlocus(seeref- theparthenogenesisallelesdidnotexpressthetraitunlessdiplo-
erencesinOzias-Akins2006).Someofthesestudieshaveshown sporywasalsopresent,apossibilityalsosuggestedbythework
that the region controlling aposporous apomixis has reduced ofAlbertinietal.(2001)inP.pratensis.However,themostrecent
recombination,withmultiplemarkersandpotentiallymultiple work on Erigeron (Noyes et al. 2007) has found autonomous
genesfallingwithinthelinkagegroup(Ozias-Akins2006).The development of embryos and endosperm in Erigeron without
onlyclearexampleofsegregationofaposporyfrompartheno- diplospory; aneuploidy may be responsible for the incomplete
genesisisinPoapratensis(Albertinietal.2001;Matzketal. penetrance observed in the earlier work. Even so, partheno-
2004). In this case, whereas apospory and parthenogenesis genesis was more extensive when diplospory was also
WHITTON ETAL.—APOMIXIS IN ANGIOSPERMS 173
present.Thesestudiessuggestthatitispossiblethatasexual etes would survive for long within a sexual population. In a
diploid could harbor alleles for parthenogenesis with little to sexualdiploid,notonlywouldsuchamutationgenerateahigh
nopenetranceandthattheexpressionofsuchallelescouldin- frequency of triploid offspring, but also pollen fitness would
creaseonceunreducedovulesareproduced. beseverelyreducedifallelesfortheproductionofunreduced
Anotherpossibleroute tothe coestablishmentofunreduced megagametophytesarelethalinhaploidgametes(Nogler1984;
embryoproductionandparthenogenesisisthroughhybridiza- Rocheetal.2001).
tion, a long-recognized correlate of apomixis (Stebbins 1950; Another possible causal link between unreduced gametes
Asker and Jerling 1992). In a number of examples, hybrids and polyploidy assumes that apomixis first becomes estab-
produceamuchhigherfrequencyofunreducedgametesthan lishedindiploidsbyasimultaneousincreaseinthefrequency
do their parents (Ramsey and Schemske 1998, their supple- of unreduced gametes and parthenogenesis (as discussed pre-
mentarytable1),withfrequenciesreaching86%.Thissuggests viously). In such a nascent apomictic population, occasional
thatmeiosisisderegulatedtosomeextentinnewlyformedhy- fertilization of unreduced eggs may occur, generating higher
brids,leadingtoanautomaticincreaseinunreducedgametes. ploidy levels until suchtimethat strictermechanismstoavoid
Simultaneously, parthenogenesis might be favored in hybrids fertilization have evolved, which would produce the associa-
toavoidtheproductionoflower-fitnessF offspringresulting tion between apomixis and polyploidy as well as patterns of
2
fromgeneticincompatibilities.Ifunreducedgameteformation multipleploidylevelsobservedinmanyapomicticgroups.Al-
in hybrids leads to triploid formation, then later-generation thoughthismodeldoesnotprecludethepersistenceofdiploid
hybrids could also have decreased fitness due to segregation apomicts,onewouldpredictthatploidylevelswouldrisebya
problems, further favoring parthenogenesis. In addition, re- ratchet mechanism (Meyers and Levin 2006) if increases in
currenthybridizationalsoprovidesacontinualsupplyofindi- ploidy are much more common than reductions. Indeed, the
vidualsofdifferentgeneticconstitutions,increasingthechance modelingresultsofMeyersandLevin(2006)suggestthatbe-
thatalineagecapableofbothunreducedgameteformationand cause of the irreversibility of increases in ploidy level, poly-
parthenogenesisarises. ploids can achieve high frequencies within lineages over time
Anotherobservationofinterestisthattheallelesorlinkage even if they are at a selective disadvantage. In a system with
blocks contributing to apomixis are, at least in some cases, apomixis,diploidapomictswouldbesubjecttostochasticloss,
lethalinhaploidpollen(Nogler1984;Rocheetal.2001).Be- withlittlepossibilityofreplenishmentthroughploidyreduction.
causegametophyticapomixisistypicallyassociatedwithreg- Inaddition,anyfactorthatlimitedthepersistenceorspreadof
ular meiosis in pollen (Mogie 1992; see recent examples in apomixisgenesindiploidswouldaccentuatethispattern.
Bicknelletal.2003;Noyesetal.2007;TalentandDickinson Itremainspossible,however,thatthelinkbetweenpolyploidy
2007), diploid carriers would be unable to pass such alleles andapomixisisnotcausallyrelatedtounreducedgametefor-
totheiroffspringthroughpollen.Thissuggeststhatthemuta- mation. Rather, apomixis and polyploidy might be induced
tions leading to apomixis likely established in lineages either by yet another factor. For example, both apomixis and poly-
thathadpreviouslybecome polyploid(sothat themutantal- ploidization could be selectively favored in hybrids between
lelegainedfitnessthroughpollen) orthat werepredisposed to two species as mechanisms to avoid hybrid breakdown and
parthenogeneticdevelopment(sothatthemutantallelegained toensurebalancedsegregationofparentalgenomes.Suchpoly-
fitness through apomictic seeds). This assumes that haploid ploids,originatingfromcrossesbetweendifferentspecies,are
lethality is associated with apomixis from the beginning. Al- known as allopolyploids, whereas those polyploids originat-
ternatively,haploidpollenofapomictsmightbeviableinitially ing within one species are termed autopolyploids. Autopoly-
andthenmightgraduallydeteriorateduetotherelaxationof ploidy has traditionally been thought to contribute rarely to
selectionpressureonpollen. long-lastingpolyploidlineages,butthishasrecentlybeencalled
into question. Soltis et al. (2007) point out several cases in
which autopolyploid origins are more parsimonious. Deter-
The Expression of Apomixis in Polyploids mining whether there is a bias toward allopolyploidy among
apomicts will help to shed light on the causal mechanisms
These findings also shed some light on the association be- linkingapomixisandpolyploidy.Ifmostapomictsareallopoly-
tween gametophytic apomixis and polyploidy. Because the ploid,wewouldinferthathybridizationisthekeycausallink
formationofunreducedmegagametophytesisacharacteristic betweenapomixisandpolyploidy.Ifallopolyploidyisnotover-
featureofgametophyticapomicts,aswellasakeyelementin representedamongapomicts,itsuggeststhateitherpolyploidy
the production of polyploids in sexuals (Harlan and deWet predatedapomixisorapomixispredatedthecompleteexclusion
1975; Ramsey and Schemske 1998), it is tempting to infer ofpollen(seeabove).
thatunreducedeggformationisthecausallinkbetweenapo-
mixisandpolyploidy.However,whetherthisisthecausallink
andexactlyhowitmightworkremainunclear. Phylogenetic Patterns in the Origin and
Onepossibilityisthatamutationcausingahighfrequency Spread of Apomixis
of unreduced gametes first becomes established within a sex-
ualpopulation,generatingahighfrequencyofpolyploidsbut It is difficult to determine the rate of appearance of apo-
also selecting for parthenogenetic development to avoid low- micticlineagesviamutationbecauseapomixiscanalsospread
fitness offspring of even higher ploidy levels. However, one viahybridization.Apomixisoftenappearsrepeatedlyinclosely
problem with this model is that it is difficult to explain how relatedspeciesassemblagesknownasagamiccomplexes(Grant
a mutation generating a high frequency of unreduced gam- 1981). This pattern could indicate that apomixis arises more
174 INTERNATIONAL JOURNAL OF PLANTSCIENCES
readilyin somegroupsorthatitspreads moreeasilyin these neticaffinitiesoftheagamospeciesC.barbigerausingvariation
groups.OurworkintwoNorthAmericanagamiccomplexes inchloroplastDNA.
in the Asteraceae is used here to illustrate the challenges in Crepis barbigera is most common in the dry regions of
interpretingtheoriginsandspreadofapomixisinsystemswith central and eastern Washington and occurs less frequently in
complexphylogenetichistories. adjacentIdahoandOregon.Acrossitsrange,itoverlapswith
TheNorthAmericanCrepisagamiccomplexincludesseven a number of other species in the agamic complex. The most
taxonomic species and more than 100 apomictic forms. This common taxa occurring within the range of C. barbigera are
groupwasstudiedextensivelyinthe1920sand1930sbyBabcock Crepis acuminata, Crepis atribarba, and Crepis modocensis
and Stebbins, who conducted morphological and cytological ssp.rostrata.BabcockandStebbinshypothesizedthatC.bar-
surveystocharacterizepatternsofvariation.Sevenofthespe- bigera is a complex derivative of these three taxa. Because
cies include both sexual diploids and polyploids that can be C. intermedia, Crepis occidentalis, and C. modocensis spp.
sexualbutaremoreoftenapomictic.Althoughmostpolyploid modocensis are also sympatric with these taxa and overlap
apomictsareassignedtooneoftheseseventaxonomicspecies morphologically,wealsoincludedthesespeciesinourstudy.
alongwithdiploidsexuals,polyploidsoftencombinecharacter- Three strongly supported nodes emerge from our analysis
istics of two or more diploids. The remaining two species, (fig. 1). We find that populations identified as C. barbigera
CrepisbarbigeraandCrepisintermedia,aretermed‘‘agamospe- (following the keys of Babcock and Stebbins 1938) occur in
cies,’’knownonlyaspolyploidapomictswithdistinctmorpho- at least two distinct clusters, one northern (Washington) and
logical characteristics. All taxonomic species in the complex one southern (Oregon), with populations of the other taxa,
are connected through an intricate web of hybrid apomictic including C. atribarba and C. intermedia, intermingled within
polyploids, and no part of the complex is completely geneti- thesetwoclades.PopulationsofC.modocensisssp.modocensis
cally isolated from the rest (Babcock and Stebbins 1938; andC.occidentalisoccupyasomewhatdistinctpositiononthese
Whitton1994).Weareusingphylogeneticanalysesofpopula- trees.Althoughthereistaxonomicuncertaintyaboutwhether
tionswithintheagamiccomplextoinferevolutionaryrelation- northern and southern populations of C. barbigera should
ships among members of the complex, to clarify taxonomic be treated as conspecific (C. Sears, personal communication),
treatments,andtoshedlightontheoriginsandspreadofapo- the general lack of correspondence between the chloroplast
mixis.Here,wedescriberesultsofastudyexaminingphyloge- phylogeny and the morphological species identity suggest that
Fig.1 Unrootedstrictconsensusoftreefrom120maximumparsimonyanalysisoffourcpDNAregionsforsamplesoftheNorthAmerican
CrepiscomplexfromWashingtonandOregon.Bootstrapsupportvaluesofmorethan50%areindicatedabovethenodes.Sampledindividuals
representsevenpopulationsofCrepisbarbigera,twopopulationsofCrepisacuminata,fourpopulationsofCrepisatribarba,twopopulationsof
Crepismodocensis,fourpopulationsofCrepisintermedia,andtwopopulationsofCrepisoccidentalis.GenomicDNAwasisolatedfromfield-
collectedleafsamplesdriedinsilicagelusingamodifiedCTABprotocol(Michielsetal.2003).FourchloroplastDNAregions(trnG,Rpl16,
rps16,andtrnM-trns;Shawetal.2005),coveringca.3507bp,weresequencedinbothdirectionsusingBigDyechemistryandwererunonan
AppliedBiosystems3730S48-capillarysequencerattheUniversityofBritishColumbiaNucleicAcidProteinServiceunit.Rawsequenceswere
proofreadandeditedinSequencher,version4.2.2,andweremanuallyalignedinSe-Al,version2.0a(Rambaut1996).Nucleotidesubstitutions
andunambiguousindelswereusedascharactersinphylogeneticanalysis.Thirty-fivephylogeneticallyinformativecharacterswerepresentinthe
combineddataset.ThedatawereanalyzedusingPAUP,version4.0b(Swofford2003),underthemaximumparsimonycriterionwithheuristic
search,using1000replicatesandtree-bisection-branchswappingonthebesttreesonly.
WHITTON ETAL.—APOMIXIS IN ANGIOSPERMS 175
polyploidapomictsinCrepishaveacomplexhistoryofhybridi- 1989]andTownsendia[Thompsonetal.2008])findthatsin-
zation,inagreementwith previous studies(Babcockand Steb- gleclonestendtobegeographicallyrestrictedtooneorasmall
bins1938;Holsingeretal.1999).Theexistenceofgeneflow number of populations (summarized originally in Ellstrand
within the complex suggeststhat apomixismighthavespread andRoose1987).Furtherstudiesareneededtoconfirmthese
viahybridization.Itis also possible that diploid hybrids have findings, which suggest that apomictic lineages rarely become
increased frequencies of unreduced gamete formation and widespreadthroughseeddispersalalone.
that this contributes to both increases in ploidy level and en- That asexuals do not appear to have a major advantage
hanced expression of apomixis. Although seemingly less parsi- throughproliferationsuggeststheneedtoreexamineassump-
monious, it is also possible that apomixis has arisen de novo tions regarding the costs of sex. If an asexual mutant makes
multipletimes.Ourunderstandingwillbegreatlyenhancedif noinvestmentinmalefunctionandisabletoreallocatethese
we can follow the evolutionary history of genes underlying resources to seed production, then the asexual lineage is pre-
apomixis.Insuchaninterconnectedreticulatingcomplex,we dicted to have a direct twofold transmission advantage. This
wouldpredictthatapomixisalleleshavespreadextensively. assumes that competing sexual species allocate reproductive
IncontrasttothesituationinCrepis,apomixisinTownsendia resourcesequallytomaleandfemalefunctions,asispredicted
occursintheabsenceofevidenceofextensivehybridization,but tooccuramongoutcrossingspecies.Onefactorthatcouldre-
recent worksuggests that apomixis can alsobecome repeatedly ducethecostofsexpaidbyrelatedsexualsisselfing,asallo-
establishedinthisgroup.InTownsendia,apomixisoccursinap- cationtomalefunctionispredictedandobservedtobelower
proximatelyhalfoftheca.30taxa,whereitisrestrictedtotet- among selfers (Ritland and Ritland 1989; Parachnowitsch
raploids. All polyploids are described as autopolyploids using and Elle 2004). However, apomixis tends to arise from self-
morphologicalcriteria(Beaman1957).Autopolyploidoriginsof incompatible progenitors (Asker and Jerling 1992; see Roy
apomicts are indicated in Townsendia hookeri (Thompson and 1995 for a notable exception in Boechera [formerly Arabis]
Whitton 2006) because there is no evidence of morphologi- holboellii), which are likely to suffer from the full cost of
cal hybridity in diploid or polyploid populations. All sampled sex.Of course,apomicts willgain benefits from avoiding the
T. hookeri polyploids have chloroplast haplotypes that either costofsexonlyifthereallocationtofemalefunctionaccom-
areidenticaltoorrecentlyderivedfromthoseofdiploidconspe- panies the appearance of apomixis (Maynard Smith 1978;
cifics.PhylogeneticanalysisofT.hookerichloroplastDNAsug- Mogie1992).Becausemutationsthatproducemeioticabnor-
gests a minimum of four evolutionary transitions from sexual malitiesinovulesoftendonotaffectpollendevelopment,the
diploidy to apomictic polyploidy (Thompson and Whitton transitiontoapomixisneednotreducepolleninvestment.This
2006),butitseemsmoreplausible thattheallelesforapomixis suggeststhattheinitialspreadofapomixismaybelittleaided
spreadintothesefourlineagesratherthaninvokingfourdistinct by reallocating resources away from male function. Empirical
origins of apomixis alleles. Alternatively, apomixis alleles may studiesexaminingtheimmediateconsequencesformalefertility
representasharedancestralpolymorphismthatgainsexpression areneededinnovelapomictstoquantifytheextentto which
inpolyploidpopulationsofT.hookeri(andpossiblyotherapo- theypaylessofacostofsex.
mictic townsendias). It is difficult to explain how apomixis al- Overthelongerterm,asinvestmentinpollendeclines,apo-
lelescouldpersistindiploids,unlesstheexpressionofapomixis micticlineagesshouldbegintoreapthefullbenefitsofavoid-
issomehowploidydependent.Onceagain,wefindthataclear ingthecostsofsex.Twostudiesdemonstratethatestablished
understandingoftraitevolutionwillbegreatlyadvancedbythe apomicts may gain a nearly twofold reproductive advantage.
abilitytotrackthehistoryofapomixisalleles. Meirmans et al. (2006) compared allocation patterns of nat-
urally occurring male sterile and male fertile apomictic dan-
delions. They found significantly greater (but not twofold)
Pathways for the Spread of Apomixis reallocationintheformoffruitnumberinmalesterileplants.
O’Connell and Eckert (2001) provide evidence of a twofold
Giventhatanapomictariseswithinasexualpopulation,there advantage in apomictic females of the dioecious Antennaria
aretwoprimarywaysforasexualitytospreadoverspace:direct parlinii.Wethusexpectatransitiontooccuraspolleninvest-
dispersalviaapomicticseedandindirecttransmissionviapol- mentdeclines,withyoungapomicticlineageshavinglittlefer-
len. In addition, apomixis genes can spread via any sexual tilityadvantageoverrelatedsexualpopulations,whereasolder
seeds produced by an apomictic parent and through vegeta- apomictsgainfromareducedcostofsex.
tivepropagation.Vegetativespreadtendstobegeographically
circumscribed and is less likely to differ between apomicts
Indirect Transmission via Pollen
andrelatedsexuals(butseeO’ConnellandEckert1999).
The previous discussion assumes that pollen production
onlyslows the spread of apomixis, butthis need notbe true.
Direct Dispersal
Ironically,thegenesformaternalclonalitycanbetransmitted
Thedirectdispersalofapomicticprogenyisthemostobvi- viamalegametes,andthismodeoftransmissionmaywellbe
ous and straightforward means by which the trait could important in the establishment and spread of apomixis. That
spreadoverspace.Indeed,itiscommonlyassumedthatasex- the genes for apomixis can be transferred to sexuals via the
ual plants should spread at an even higher rate than sexual pollenofapomictshaslongbeenknown.Experimentalcrosses
relatives because they avoidthe costs of sex (Maynard Smith have made use of the pollen of apomicts beginning in the
1978).However,thisassumptionmaynotbevalidassurveys 1940s(Asker and Jerling 1992; e.g.,Tas and Van Dijk 1999;
conducted in a number of apomicts (e.g., Antennaria [Bayer Brock 2004), and this has become the prevalent means of
176 INTERNATIONAL JOURNAL OF PLANTSCIENCES
studying the inheritance of apomixis (Noyes and Rieseberg (comparedwith50%inobligateoutcrossers)aswellasbeing
2000). 50%relatedtotheseedsthattheysireviapollen.Ifpollensuc-
Althoughindividualcrossingstudiesprovideproofofprin- cessissimilaramongapomictsandsexuals,thiswouldgenerate
ciple for the inheritance of apomixis via the pollen of apo- a3:2transmissionadvantagethatparallelsthetransmission
micts, studies of wild-collected individuals are just beginning advantageenjoyedbyselfersarisingwithinoutcrossingpopu-
to uncover the importance of this mechanism in nature. Sur- lations (Maynard Smith 1978; Uyenoyama 1984; Holsinger
veys of apomictic populations have documented patterns of 2000).Themagnitudeofthetransmissionadvantagewillde-
isozymeandcpDNAvariationconsistentwithmultipleorigins pendonlevelsofpollenviabilityinnewapomicts,aboutwhich
of apomixis or transmission of apomixis via pollen (Roy and little is known. In established apomictic populations of T.
Rieseberg 1989; Sharbel and Mitchell-Olds 2001; Thompson hookeri, pollen viability averaged 17%, with the highest via-
and Whitton 2006). However, so far thesedata do notallow bility being 42% (Thompson and Whitton 2006). However,
assessment of the frequency of transmission of apomixis via Noyesetal.(2007)producedanovelapomictwith98%via-
pollen. blepollen.
The transmission of apomixis genes to sexuals via pollen MaynardSmith(1978,p.41)notedthatasthefrequencyof
may be of long-term importance for the spread of apomixis, apomixisrises, transmission via pollen becomesless efficient,
eveniftheseeventsarerelativelyinfrequent.Thus,althoughit andthus,onewouldexpectpollenfertilitytodecline(through
isstraightforwardtoexaminethereproductivemodeofopen- driftorselection;MaynardSmith1978;Eckert2002).Because
pollinatedseedofsexualsforevidenceoftransmissionofapo- apomixiswouldnotenjoyatransmission advantagewithina
mixis,toolsthatfacilitatescreeningoflargenumbersofprogeny highlyselfingpopulation(inwhichoutcrosspollenisrarelyin-
maybenecessarytoallowdetectionofrareeventsanddocu- corporated),theobservationthatapomixistypicallyarisesfrom
ment this pathway in nature. Two recently developed ap- self-incompatible species (Asker and Jerling 1992) suggests
proaches could prove essential in understanding the role of thattheinitialtransmissionadvantagearisingfromproducing
pollen in the transmission of apomixis. Flow cytometric seed pollen as well as apomictic seeds might be important to the
screening(FCSS;Matzketal.2000)canbeusedtoassessthe spreadofapomixis.Interestingly,apomictsalsotendtomain-
relativeploidyofendospermandembryoandthustodetermine tainpollinatorattractionfeaturesandarenotassociatedwith
thegeneticoriginofseeds.Forexample,atetraploidautono- reducedflowersize,againsuggestingthatpollentransmission
mousapomictisexpectedtohavea4:8embryo :endosperm continuestobeimportantinthesetaxa.
ploidyratio(assumingthattwounreducedpolarnucleifuseto Whenpollenfromapomictsfertilizessexualovules,thisallows
formtheendosperm),whileasexualtetraploidwouldhavea the genes for apomixis to move into multiple genetic back-
4:6ratio(assumingthattworeducedpolar nucleifusewith grounds; this is perhaps even more important than the trans-
a reduced pollen nucleus). To date, the use of this approach mission advantage gained. As pointed out by Mogie (1992),
hasfocusedmainlyonestimatingtherateofsexualandasex- this movement might be key to the evolutionary success of
ual reproduction in apomicts (Matzk et al. 2000), but these apomicts.Geneflowviapollenimprovesthechancethatapo-
data can also indicate that pollen from apomicts sires seed mixis will be associated with a genotype of high fitness, in-
(TalentandDickinson2007).FCSScanbeappliedtonatural creasingtheprobabilityofsuccessfulestablishment.Thespread
ortoexperimentalpopulationsofmixedsexualsandasexuals. ofapomixisviapollenalsoincreasesthegeneticvariabilityof
Ma´rtonfiova´ (2006)appliedFCSStoexperimentallyproduced asexualpopulations,improvingtheirabilitytorespondtose-
progenies,involvingcrosseswithsexualdiploidsandapomic- lectionandtoadapttonewhabitats.Indeed,patternsandlevels
tic triploids and tetraploids in Taraxacum sect. Ruderalia. If of diversity within and among apomictic populations reveal
triploids were used as pollen parents, all 11 diploid mothers substantial variation (Ellstrand and Roose 1987; Silvertown
examinedproducedsometriploidprogeny.However,noneof 2008).Thus,theoccasionalcaptureofdifferentsexualgenetic
the 35 open-pollinated diploid mothers produced triploid backgroundsviapollenflowfromapomictsmayprovidemany
seed.Thissuggeststhatdetectionofgeneflowbetweensexual ofthe evolutionarybenefits of sexandrecombination topre-
diploidsandapomicticpolyploidsinnaturemayrequireinten- dominantly asexual plant populations (further discussed in
sivesampling.Oneunfortunatelimitationofthisapproachis ‘‘ApomixisandtheEvolutionofSex’’).
that seeds are sampled destructively, precluding later evalua- Apomixisgenesmayalsobetransmittedviasexuallyfertil-
tionofthereproductivemodeoftheoffspring.Inanalterna- izedseedsproducedbyotherwiseapomicticplants(Askerand
tive approach, Bicknell et al. (2003) made use of transgenic Jerling1992;Matzketal.2000;Bicknelletal.2003;Richards
Hieracium pilosella individuals carrying antibiotic resistance 2003). If this process involves meiotically reduced eggs, the
to track the reproductive dynamics of two facultatively apo- apomict would lose its transmission advantage with respect
micticindividuals.Althoughthisapproachislimitedtosystems to these seeds (which bear only half of the genes of the ma-
in which transformation protocols have been worked out, it ternal apomictic parent). If unreduced eggs were involved,
allowsscreeningoflargenumbersofseedsandefficientrecov- the apomict would continue to transmit the same number
eryofraresexualoffspringunderexperimentalconditions. ofgenestothenextgenerationviatheseseeds,buttheseeds
Totheextentthat pollenfromapomictsfertilizes ovulesin would have an increased ploidy level and might well be less
sexualindividualsandproducesapomicticoffspring,apomictic fit.Thus,wewouldexpectalowertransmissionadvantagein
lineageswillgainatransmissionadvantageeveniftheydonot apomicts that often act as maternal parents of sexually pro-
reallocate resources from male to female function (Maynard duced seed. However, to the extent that this pathway allows
Smith1978;Holsinger2000;BrittonandMogie2001).Apo- apomixisto be expressed in novelgenetic backgrounds, ittoo
mictsare100%relatedtotheseedsthattheyproducedirectly mayincreasethelikelihoodofpersistence.
WHITTON ETAL.—APOMIXIS IN ANGIOSPERMS 177
Geographical Parthenogenesis strong reproductive incompatibilities through the formation
ofinviableorsterileoffspring(e.g.,triploidblock;Ramseyand
Thatsexualandasexualrelativestendtohavenonoverlap- Schemske 1998; Burton and Husband 2000). Where sexual
ping ranges in both plants and animals has long been noted, diploidsandasexualpolyploidsco-occur,hybridizationasym-
resulting in a pattern termed ‘‘geographical parthenogenesis’’ metrically reduces the fitness of sexual diploids because sex-
by Vandel (1928). The distributions of a number of taxa re- ualsaremorelikelythanapomictstoincorporatepollenfrom
vealthatasexualsarelikelytooccupylatitudesandelevations other ploidy levels, leading to offspring of odd ploidy levels.
higherthanthoseoccupiedbytheirsexualcounterparts.Data Thus,allelsebeingequal,thepresenceofasexualscandesta-
relating to geographical parthenogenesis in plants are sum- bilizesexualpopulations.Sexualsmightthereforepersistonly
marizedbyBierzychudek(1985)andHo¨randl(2006). inportionsoftherangewheretheyhaveaselectiveadvantage
Explanations for the existence of geographical partheno- overasexuals.Inthisregard,itisnoteworthythatinbothCrepis
genesis generally fall into either ecological or demographic and Townsendia, diploids tend to occupy habitats that could
categories. Ecological explanations are based on differential bedescribedasmorespecialized.Forexample,inT.hookeri,
ecological tolerances or competitive abilities of sexuals and diploids occur on Niobrara and Pierre shales, with soils that
asexuals.Forexample,ithasbeenproposedthatasexualsmay aresignificantlyhigherinselenium,aluminum,andotherele-
comprisegeneral-purposegenotypes(Lynch1984)thatcanoc- ments known to restrict the distribution of plants in other
cupy a wide range of ecological conditions but are relatively groups(E.M.StaceyandJ.Whitton,unpublisheddata).The
poor at tracking extreme or specialized habitats. Support for restrictionofdiploidstotheseunusualsubstratesmayindicate
theexistenceofgeneral-purposegenotypesinplantsis,however, thattheyholdaselectiveadvantageundertheseconditions.
sparse. Although it might seem unlikely that multiple geneti-
cally distinct lineages of apomicts would have hit on similar
general-purpose genotypes, independent lineages of apomicts Apomixis and the Evolution of Sex
mightexhibitsimilarecologicalfeaturesasaresultofparallel
polyploidizationevents;polyploidyofteninducesspecificmor- One of the longest-standing questions in evolutionary biol-
phological changes, altered ecological tolerances,and shifts in ogyiswhysexissoprevalent.Thatapomixisarisesrepeatedly
geneexpression(OttoandWhitton2000;AdamsandWendel inplantsbutfails totake substantialholdsuggeststhatplants
2005).Weareawareofnostudiestodatethatexaminetheeco- mayshedlightonthebenefitsofsexandthecostsofasexual-
logicalresponseofmultiplelineagesofasingleapomict. ity. Here, we discuss how the emerging data on apomixis in-
Demographicexplanationsattributethedistributionofsex- formstheliteratureontheevolutionofsexandviceversa.
uals and asexuals to the relative dispersal and establishment To structure this discussion, we focus on three interrelated
abilities of apomicts. In addition to the ability to found pop- questions:Whyaretherenoancientasexualangiosperms?Why
ulationswithasinglepropagule,apomictsmaybefavoreddur- do few plant populations exhibit mixtures of sexual and asex-
ing recolonization of previously glaciated habitats because of ualindividuals?Andwhydon’tmostplantsreproduceapomic-
the scarcity of potential pollen sources and pollinators. We ticallywhileengaginginoccasionalsex?
would thus predict that apomicts that colonize deglaciated
areaswouldmoreoftenbeautonomousorbeabletouseself
Why Are There No Ancient Asexual Angiosperms?
pollen to initiate endosperm. Once established in these habi-
tats,apomicticplantscouldexcludesexual relatives. owingto Although apomixis may appear fixed or nearly so within
frequency-dependentfitnesseffectsdescribedbelow. particularpopulationsorportionsofaspeciesrange,theonly
Becauseinplantsgeographicalparthenogenesistendstoin- knowninstance of aphylogenetically isolated apomicticline-
volvegametophyticapomixisandbecausesuchapomictstend ageinseedplantsisHouttuynia,asmall(perhapsmonotypic)
to be polyploid (unlike with adventitious embryony), Bierzy- genus in Saururaceae, first reported as apomictic by Okabe
chudek(1985)suggestedthatpolyploidyratherthanasexual- (1930).However,itisnotclearthattheentirelineageisapo-
ity may be the key contributor to the distributional patterns. mictic.Evenifitis,whetherthelossofsexoccurred earlyor
Stebbins(1950)alsoobservedthatsexualpolyploidstendtohave lateinitsevolutionaryhistoryisnotknown.Allremainingan-
morenortherlyandhigher-elevationdistributionsthanrelated giosperm apomicts are at the tips of the tree of life, with no
diploidsexuals.Brochmannetal.(2004) confirmedthatpoly- other higher-level taxa (families, genera) that are asexual. A
ploidsaremoreprevalentinthearcticbutnotedthatthistrend number of well-studied temperate apomicts have geographic
appliedmainlytohigherploidylevels(greaterthantetraploid) distributions consistent with an origin following Pleistocene
andtoflorasinpreviouslyglaciatedregions.Morefrequently, glaciation, suggesting that these apomicts may be of very re-
authorssuggestthatasexualityratherthanpolyploidyis key in cent origin (Ho¨randl 2006; Thompson and Whitton 2006).
establishingthepattern(Ho¨randl2006;ThompsonandWhitton Despite this phylogenetic distribution, there are indications
2006).Therelativecontributionsofapomixisandpolyploidy that the tendency to produce apomictic lineages may have
togeographicpatternswillbebestexploredinsystemsinwhich arisen early in the diversification of some deeper lineages. For
both sexual and apomictic polyploids are known, such as in example,apomixisintheRosaceaeoccurswithinspeciesofnu-
theAntennariaparliniicomplex(BayerandStebbins1983). merous genera. Recent phylogenetic results (Potter et al. 2007)
Polyploidyandapomixismayalsointeractinamannerthat reveal that Rubus, which contains a number of apomicts, di-
contributes to geographical parthenogenesis, through a phe- vergedearlyintheevolutionaryhistoryofsubfamilyRosoideae.
nomenonknownasdestabilizinghybridization(Lynch1984). This suggests that the tendency togenerate apomicts may have
Interploidycrossesinsexuallyreproducinggroupscangenerate appeared early within this lineage. Although a predisposition
178 INTERNATIONAL JOURNAL OF PLANTSCIENCES
towardapomixismightpersistforlongperiodsofevolutionary tionsisnotreplenishedathighrates by migration(Thompson
time, apomictic lineages, once produced, have only a shallow andWhitton2006;Thompsonetal.2008).
phylogenetic history in seed plants, with no truly ancient asex-
ualtaxa.
Why Do Few Plant Populations Exhibit Mixtures
Why apomictic lineages of angiosperms appear to be re-
of Sexual and Asexual Individuals?
strictedtotheverytipsofevolutionarybranchesremainsun-
clear. Ancient asexual lineages appear in other groups, most Characterization of reproductive mode in apomictic com-
notably bdelloid rotifers and darwinulid ostracods but also plexes suggests that it is uncommon for local populations to
amongferns, wherevegetativelyreproducingVittariasp.and becomposedofsubstantialnumbersofbothprimarilysexual
Trichomanes sp. are estimated to be ca. 10 million years old and primarily apomictic individuals (Beaman 1957; Gadella
(Farrar1990). 1987; Asker and Jerling 1992; O’Connell and Eckert 1999).
Recent theoretical work on the evolution of sex has high- Whereco-occurrencehasbeendocumented,itoftenresultsfrom
lightedtheimportanceofmaintainingsexandrecombination secondary contact between sexuals and apomicts, producing
inpopulationsthatarelimitedinsizebecausegeneticvariation complexsexual-asexualdynamics(Verduijnetal.2004;Noyes
can be rapidly depleted by the combined action of drift and and Allison 2005; Thompson et al. 2008; K. M. Dlugosch,
selection(seereviewsinOttoandGerstein2006;deVisserand C.J.Sears,andJ.Whitton,unpublishedmanuscript).Anotable
Elena2007).Geneticvariationinfitnessisparticularlylimited exception occurs in Taraxacum, in which ploidy cycles, with
in asexuals because of their inability to bring together high- accompanyingswitchesbetweensexualityandapomixis,seem
fitnessallelesthatresideindifferentindividuals.Consequently, toariseonshorter timescales (Meirmansetal.2003;Verduijn
anyadvantagethatapomictsmightinitiallyhaveisthoughtto etal.2004),basedonstudiesofstandingvariation.
decrease over timebecause of(a) theaccumulation of delete- That populations tend to be predominantly apomictic or
rious mutations (Muller 1932; Keightley andOtto 2006), (b) predominantlysexualisconsistentwithmodelsexploringthe
a reduction in the rate of adaptation due to beneficial muta- outcome of competition between asexuals and sexuals. In a
tionsthatariseindifferentindividualsandcannotbebrought numberofecologicalmodels,itisassumedthatwhateverad-
together in apomicts (the Fisher-Muller hypothesis; Morgan vantage(s) sex has in the long run can be modeled as a fixed
1913;Fisher1930;Muller 1932;Barton andOtto2005), and fitness advantage (Joshi and Moody 1995, 1998; Bengtsson
(c) a greater loss of beneficial mutations because they happen and Ceplitis 2000; Britton and Mogie 2001; Carrillo et al.
toariseingenomescarryingdeleteriousmutations(theruby-in- 2002).Under theseassumptions, whether apomictsspreadto
the-rubbishhypothesis;Fisher1930;Peck1994).Ineachcase, fixationdependsontheassumedfitnessadvantageofsexuals,
the fortune ofapomicts ispredictedtodecline over evolution- therelativepollenoutputoftheapomict,theabilityofapomicts
arytimescalesbecauseofthedifficultyofcombiningfitalleles. to reallocate male function, and the relative success of male
In this context, there are two potential ways for obligate gametesfromapomicts(JoshiandMoody1995).Inbothspa-
asexual species to avoid extinction. First, an asexual lineage tial (Britton and Mogie 2001; Carrillo et al. 2002) and non-
couldavoidtheaccumulationofdeleteriousmutationsifmu- spatial (Joshi and Moody 1995; Bengtsson and Ceplitis 2000)
tationrateshappentobeparticularlylow.Suchalineagewould versionsofthesemodels,itisdifficulttomaintainsexualsand
be unable to adapt rapidly, but if its environment were suffi- asexualstogetherindefinitely.Thus,thesemodelspredictthat
cientlystable,itmightavoidextinction.Somedatasuggestthat either sexuals or asexuals will win out within a population,
darwinulidostracodsmighthavehitonthislong-termsolution dependingontheexactbalanceofselectiveforces.
because ostracods have unusually low mutation rates (Butlin These models assume that the benefits of sex are fixed,
and Menozzi2000; Scho¨net al. 2003).Second, ifanasexual whichisunlikelytobetrue.Forexample,newlyformedapo-
specieshasasufficientlylargepopulationsizeandahighmi- micts may suffer little or no loss of viability compared with
grationrateamongpopulations,geneticvariationwithinlocal sexuals because there has been little time for mutation accu-
populationscouldremainhighenoughtoallowasexualsto mulation. In contrast to these ecological models, evolution-
evolveatarateakintothatofsexualcompetitors(Ladleetal. ary models track the dynamics of selected loci in sexual and
1993; Judson 1995; Martin et al. 2006; Salathe´ et al. 2006). asexual populations and allow the benefits of sex to emerge
This is perhaps the route taken by bdelloid rotifers because from differences in the response to selection. Nevertheless,
analysesofsequencedatasuggestthatdispersalratesarehigh, evolutionarymodelsalsotypicallyobservethateithersexuality
genetic diversity levels are similar to those of sexual popula- or asexuality comes to dominate a population. For example,
tions,andtheabilitytorespondtoselectionissimilartothat spatiallyexplicitmodelsofindividualssubjecttorecurrentdel-
ofsexualmonogonontrotifers(Birkyetal.2005). eterious mutation find that asexuals beat out sexuals when
So why aren’t there ancient asexual plants? Perhaps the dispersal rates are high or mutation is low, while the reverse
specialcircumstancesenablingasexualstopersistoverlongpe- istruewhendispersalislocalandmutationiscommon(Peck
riodsofevolutionarytimedonotco-occurinplants.Estimates 1994;Salathe´ etal.2006).Similarly,inclassicalhost-parasite
obtained to date of the genome-wide deleterious mutation models, Hamilton (1980; Hamilton et al. 1990) found that
rate are high in angiosperms (typically above 0.1 per genera- sexualstookoverthepopulationundersomeconditions(many
tion; Drake et al. 1998; Schoen 2005). Furthermore, as a re- parasitesperhost;strongselection)butthatasexualsotherwise
sult of limited seed dispersal and low rates of gene flow via dominated.
pollen, strong spatial structure with isolation by distance is Althoughlocalpopulationstendtobeapomicticorsexual,
likely to exist in many apomictic species (Meirmans et al. both sexual and apomictic populations typically coexist on a
2003), suggesting that genetic variation within local popula- regionalorspecieswidescale.Thisobservationisconsistentwith